nodetool toppartitions samples cluster traffic and reports the partitions queried most often during a sampling period. Scope the result to a keyspace or table when investigating unusually heavy traffic. section 9.3.4
Use it to investigate hot partitions: a partition receiving more traffic than its shard can handle can increase latency for operations using that shard. The result helps connect the slowdown to a partition key, a recent application change, or an abusive traffic pattern. section 9.3.4 section 11.3.1
Finding a heavily queried author partition
Run `nodetool toppartitions reviews authors 10000` while investigating the `reviews.authors` table. A sample can report the partition key `33a753de-704c-11ee-b962-0242ac120002` with 39 reads, identifying it as the busiest recorded partition during sampling. section 9.3.4
(scylla-1) root@scylla-1:/# nodetool toppartitions reviews authors 10000
WRITES Sampler:
Cardinality: ~0 (256 capacity)
Top 10 partitions:
Nothing recorded during sampling period...
READS Sampler:
Cardinality: ~1 (256 capacity)
Top 10 partitions:
Partition Count +/-
(reviews:authors) 33a753de-704c-11ee-b962-0242ac120002 39 0
Inspect the command output showing separate write and read samplers, cardinality, and the top partition count. section 9.3.4
